A Pedestrian Traffic Signal Control Method by Use of Image Processing and Entropy

点击翻译按钮获取中文摘要

A video camera is installed on a building beside an intersection, and pedestrian traffic images are taken continuously on a crosswalk. A spatial distribution of pedestrians is obtained from a taken image by use of image processing. We have developed a method to quantify the spatial distribution pattern of pedestrians using entropy, and to estimate macroscopically pedestrian traffic demand.In this study, we propose a pedestrian traffic signal control method to adjust the duration of pedestrian green lights by use of image processing and entropy. This is a method to estimate the pedestrian traffic demand from an entropy value, and to optimize the pedestrian green signal duration every signal cycle according to the entropy value. By applying this method to an intersection, a redundant time of pedestrian green lights can be excluded. Therefore, we can expect especially to relieve the traffic jams extended from a scramble intersection. Entropy threshold values to switch from the green lights to the green winking when pedestrian traffic demand decreases was examined based on the image processing results and the simulation experiments. As a result, we obtained optimal threshold values, 2.5 bits at a usual crosswalk and 3.0 bits at a scramble intersection.Since the minimum cycle of traffic signal control is one second, a simple image processing and entropy measurement method is developed to complete the processing within one second.We also propose a pedestrian traffic signal control method with a safety function for the slow walking pedestrian such as elderly person and physically handicapped person) if such a pedestrian is detected by the image processing after changing pedestrian green lights to green winking, the green winking duration is extended until the pedestrian finishes crossing at the crosswalk.